To: "Peter B. Mandeville" <mandevip@uaslp.mx>
Subject: Re: [R] Influence.measures
From: Peter Dalgaard BSA <p.dalgaard@biostat.ku.dk>
Date: 24 Jun 1999 01:15:53 +0200
In-Reply-To: "Peter B. Mandeville"'s message of "Wed, 23 Jun 1999 11:00:19"
"Peter B. Mandeville" <mandevip@uaslp.mx> writes:
> It works fine up to now with the exception of Huber's data published in
> Atkinson 1985:8 (Plots, Transformations and Regression)
>
> Observation x y
> 1 -4 2.48
> 2 -3 0.73
> 3 -2 -0.04
> 4 -1 -1.44
> 5 0 -1.32
> 6 10 0.00
>
> which is not very well excepted. The command
>
> > inflm <- influence.measures(result)
>
> reports
>
> Warning: NaNs produced in function "qf"
Erm. If you trace influence.measures, the trouble spot is in
calculating qf(infmat[, k + 3], k, n - k) > 0.9 where infmat is
dfb.1_ dfb.x dffit cov.r cook.d hat
1 1.1124006 -9.559280e-01 1.46670838 0.3291085 0.520037540 0.2897436
2 0.1260689 -8.125181e-02 0.14998404 2.2175907 0.014641199 0.2358974
3 -0.0774641 3.328392e-02 -0.08431195 2.1730596 0.004693794 0.1974359
4 -0.5320433 1.143012e-01 -0.54418271 0.9999362 0.134536856 0.1743590
5 -0.4360945 5.048678e-18 -0.43609451 1.2300164 0.096271128 0.1666667
6 8.5733290 1.841847e+01 20.31604872 0.2552637 26.398591892 0.9358974
and infmat[, k + 3] is the cook.d column. So the last item is
qf(26.39,...) which obviously won't do. I strongly suspect that you
want pf rather than qf there!
-- O__ ---- Peter Dalgaard Blegdamsvej 3 c/ /'_ --- Dept. of Biostatistics 2200 Cph. N (*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918 ~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk) FAX: (+45) 35327907 -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.- r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html Send "info", "help", or "[un]subscribe" (in the "body", not the subject !) To: r-help-request@stat.math.ethz.ch _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._